On 10/21/2014 12:34 AM, Mark Andrews via RT wrote:
>
> A lock shouldn't be necessary.  Reference counting should prevent
> this.
>
Thank you for your response.

We suspect that the issue is caused by our bind-dyndb-ldap plugin that uses
the DYNDB API (not yet merged). The plugin calls dns_view_flushcache() when
reconfiguring global forwarders. However we don't call it in section secured by
isc_task_beginexclusive() and isc_task_endexclusive(). So the references seems
to be counted right, but other threads seems to be attaching to the DB while it is
being freed.

If you can confirm that this can cause issues (race conditions) I think it is OK to
close this ticket.

Regards,
-- 
Tomas Hozza
Software Engineer - EMEA ENG Developer Experience

PGP: 1D9F3C2D
Red Hat Inc. http://cz.redhat.com